Determinants of spatial resolution in retinal prostheses: understanding the gap between geometry and vision

Abstract

Retinal prosthetic systems have emerged as a promising technology for restoring partial vision in patients with severe photoreceptor degeneration. Over the past two decades, development in epiretinal, subretinal, and suprachoroidal devices has progressed substantially, with several representative platforms demonstrating the feasibility of restoring light perception, object localization, motion discrimination, and limited-form vision in clinical settings. Despite these advances, the visual acuity achieved by existing retinal prostheses remains markedly lower than predictions based on device geometry alone. Particularly, a persistent gap exists between the theoretical spatial resolution implied by electrode or pixel spacing and clinically reported visual performance. This review examines the major factors limiting effective spatial resolution in retinal prosthetic systems, with particular emphasis on the resolution gap. Collectively, available evidence indicates that while electrode spacing may be considered a critical determinant of theoretical resolution, it does not represent a direct surrogate for clinical visual acuity. Consequently, improving prosthetic vision requires denser arrays alongside selective neural recruitment, more stable electrode-retina coupling, and stimulation strategies that optimize functionally meaningful percepts. Elucidating the resolution gap and its underlying determinants will be essential for developing next-generation retinal prosthetic systems that provide more effective visual restoration.

Introduction

Visual perception relies on the precise conversion of light into neural signals within the retina and the subsequent transmission of these signals along the visual pathway to the brain. The retina itself is a highly specialized neural tissue comprising multiple layers of interconnected neurons, including photoreceptors, bipolar cells, amacrine cells, horizontal cells, and retinal ganglion cells (RGCs) [1]. Photoreceptors, consisting of rods and cones, play a critical role in detecting light and initiating the visual signaling cascade. Upon activation, these cells convert optical stimuli into electrochemical signals that are processed through retinal circuits and transmitted to the brain via the optic nerve, ultimately enabling visual perception [2]. Disruption of this phototransduction cascade via photoreceptor degeneration results in severe visual impairment or blindness.

Among the primary causes of photoreceptor degeneration are retinitis pigmentosa (RP) and age-related macular degeneration (AMD). RP comprises a group of inherited retinal disorders characterized by the progressive loss of rod and cone photoreceptors, often resulting in night blindness and gradual peripheral vision loss before eventually impairing central vision [3]. Conversely, AMD is a leading cause of vision loss in older adults and primarily targets the macular region responsible for high-acuity central vision [4]. Although several restorative strategies, including gene- and cell-based approaches, have been explored, treatment options capable of restoring meaningful vision in advanced photoreceptor degeneration remain limited [5, 6]. Crucially, even after photoreceptor loss, portions of the inner retinal circuitry, particularly bipolar cells and RGCs, can remain sufficiently preserved despite progressive remodeling, serving as potential targets for prosthetic stimulation [7]. This preservation provides a crucial biological basis for developing retinal prosthetic devices, which electrically stimulate surviving retinal neurons to restore partial visual perception [8].

Over the past two decades, significant progress has been recorded in the development of artificial vision prostheses for restoring functional vision following severe retinal degeneration. Depending on the anatomical target of stimulation, these devices are broadly categorized into several approaches, including retina, optic nerve, and visual cortex prostheses [9]. Retinal prostheses have garnered the highest attention for their relatively direct access to surviving retinal circuits and the possibility of utilizing existing retinal signal-processing pathways.

Several retinal prosthetic systems, including epiretinal, subretinal, and suprachoroidal devices, have been developed and subjected to clinical evaluation. Representative examples include epiretinal platforms such as Argus II and more recent high-density arrays; subretinal implants such as Alpha implantable micro-photodiode stimulator (IMS) and advanced micro-photodiode stimulator (AMS); and photovoltaic subretinal systems including PRIMA. While representative retinal prosthetic systems have demonstrated the feasibility of restoring partial visual perception, including light perception, object localization, and limited-form vision, the level of restored vision remains substantially below that required for tasks, such as normal reading or detailed object recognition [10–12]. For instance, clinical trials with the epiretinal Argus II system revealed that patients could detect large objects, identify motion direction, and perform basic orientation tasks [11]. Similarly, subretinal implants, such as Alpha IMS, improved visual function, enabling letter recognition and basic shape discrimination under controlled conditions [10].

Despite these promising advances, the visual acuity achieved with current retinal prostheses remains extremely limited, typically falling well below the threshold required for functional reading or detailed object recognition. For example, the 60-electrode Argus II system yields visual acuity around 20/1260, whereas subretinal systems with much denser pixel architectures, such as Alpha IMS/AMS, show slightly better but still limited acuity in selected patients [10, 11]. Reported clinical acuity remains substantially lower than predicted values based on electrode density and spacing. Although increasing electrode density may improve theoretical spatial resolution, clinical outcomes indicate a persistent gap between electrode-geometry-predicted resolution and the actual visual acuity achieved by patients.

This resolution gap represents a fundamental limitation of current retinal prosthetic systems. Although models predict higher visual resolution with smaller electrode spacing and increased electrode counts, the actual perceptual resolution achieved by patients remains substantially below the theoretical limit, reflecting a combination of biophysical and biological determinants, including current spread, variability in the electrode-neuron distance, unintended neural activation, stimulation-parameter constraints, and degeneration-associated retinal remodeling [12–15]. Elucidating the mechanisms underlying this resolution gap is critical for advancing the next generation of artificial vision technologies. Simply increasing and reducing electrode counts or sizes, respectively, may not necessarily improve functional vision if fundamental limitations in neural activation and retinal-signal processing are not addressed. Therefore, bridging the performance gap in retinal prostheses requires a comprehensive understanding of engineering constraints and biological determinants of spatial resolution.

In this review, we evaluate the factors limiting effective spatial resolution in retinal prosthetic systems, focusing on the gap between theoretical resolution and clinical visual performance. We examine how device geometry, electrode-tissue interactions, and stimulation conditions synergistically shape perceptual outcomes, demonstrating why optimizing electrode density alone fails to improve visual acuity. Overall, we provide a clearer perspective on the major barriers to higher-resolution prosthetic vision and highlight key considerations for future research and development.

Overview of current retinal prosthetic systems

Implantation approaches and system characteristics

Retinal prosthetic systems are generally classified based on the anatomical implantation site as epiretinal, subretinal, or suprachoroidal devices. Specifically, epiretinal devices sit on the inner retinal surface facing the vitreous cavity, while subretinal devices occupy the subretinal space beneath the retina, adjacent to the retinal pigment epithelium; suprachoroidal devices are implanted in the space between the choroid and sclera [13, 16]. These three implantation approaches differ significantly in electrode-tissue distance and the extent to which residual retinal circuitry can be engaged, with both placement-related factors influencing stimulation thresholds, current spread, and spatial selectivity [13, 14, 17]. The anatomical relationships are illustrated schematically in Fig. 1, which highlights the distinct implantation planes and their relative positions to the retinal layers.

Fig. 1.

Representative anatomical locations of retinal prosthetic electrodes.aMagnified schematic showing the relative placements of epiretinal, subretinal, and suprachoroidal electrodes with respect to the retinal layers. Epiretinal electrodes are positioned on the inner retinal surface near the RGC layer, subretinal electrodes are placed beneath the neural retina near the bipolar cell layer, and suprachoroidal electrodes are positioned between the choroid and sclera. b Cross-sectional view of the eye indicating the anatomical region enlarged in panel (a), including the retina, choroid, and sclera.

(Adapted from [9], with permission from Elsevier, and from [18], distributed under the terms of the Creative Commons Attribution License)

Epiretinal prostheses are positioned on the inner retinal surface, directly adjacent to the RGC layer and facing the vitreous cavity. Because these devices primarily stimulate the output neurons of the retina, they bypass most preserved intrinsic inner retinal signal processing. Nevertheless, this epiretinal configuration offers practical advantages in surgical visualization and device fixation, with the Argus II system remaining the most recognized clinical device for this approach [11, 19]. However, ganglion cell (GC)-level stimulation may also increase the likelihood of unintended passing-axon activation, which reduces spatial specificity and promotes distorted or elongated phosphene percepts [13, 17]. More recently, higher-density epiretinal systems were developed to improve spatial resolution and functional performance. A representative device is the 256-channel Intelligent Micro Implant Eye (IMIE 256), which has demonstrated promising safety and preliminary efficacy in a first-in-human trial for patients with end-stage RP [20]. Compared with earlier epiretinal devices, IMIE 256 increases electrode count to expand functional capabilities while retaining the core general principle of GC-level activation [13, 20].

Conversely, subretinal prostheses are implanted beneath the retina in the space previously occupied by degenerated photoreceptors and adjacent to the retinal pigment epithelium, placing them closer to bipolar-cell-mediated retinal processing pathways. The subretinal strategy is conceptually attractive because it may directly engage residual inner retinal circuitry rather than relying solely on direct GC stimulation [8, 9, 13]. Representative devices include the Alpha IMS/AMS and photovoltaic PRIMA system. Alpha IMS demonstrated that subretinal stimulation restores rudimentary-form vision, motion perception, and measurable visual acuity in a subset of patients [10]. PRIMA utilizes a wireless photovoltaic array, where pulsed near-infrared (NIR) light projected from external glasses is converted into localized electrical stimulation for nearby inner retinal neurons, thereby simultaneously restoring central vision and preserving residual peripheral vision [12, 21]. Despite these advantages, subretinal implantation is surgically demanding, requiring precise device positioning, with visual outcomes remaining strongly dependent on the extent of retinal remodeling and the preservation of the retained retinal network [7, 13, 17].

A third major category is the suprachoroidal approach, where the electrode array is placed outside the retina, in the potential space between the sclera and choroid. The suprachoroidal strategy is often regarded as surgically less invasive and may reduce the risk of direct retinal trauma since the implant is positioned farther from the retina [13, 17, 22]. However, the higher electrode-retina distance generally requires higher stimulation currents and tends to broaden the electrical activation spread, potentially limiting spatial selectivity and achievable resolution [13, 17, 22]. Clinical trials of second-generation suprachoroidal devices indicate that this strategy provides valuable functional vision with an acceptable safety profile, indicating that it remains an important alternative design pathway despite its intrinsic geometric disadvantages [22].

Beyond the implantation site, retinal prosthetic systems also differ in device architecture. For example, Argus II relies on an external camera and an image processor that converts the visual scene into patterned electrical stimulation delivered telemetrically [11, 19]. Conversely, Alpha IMS incorporates subretinal photodiodes that directly detect incident light within the eye, whereas PRIMA integrates external image capture and processing with an implanted photovoltaic array that converts projected NIR light into localized retinal stimulation [10, 12]. Device-level architectural differences may also influence how visual information is captured, processed, and delivered to the retina, thereby potentially governing overall functional performance [13, 17, 23]. As summarized in Table 1, representative retinal prosthetic systems differ across the implantation site and the mechanisms underlying the acquisition and delivery of visual information. The difference also manifests in the configuration of the stimulation electrode array.

Table 1.

Representative retinal prosthetic platforms and device characteristics

Device	Developer (country)	Implantation site	Stimulation modality	Array/pixel specification	References
Argus II	Second Sight Medical Products (USA)	Epiretinal	External camera and telemetry-driven electrical stimulation	60 electrodes	[11, 17, 19]
IMIE 256	Golden Eye Bionics (USA); IntelliMicro Medical (China)	Epiretinal	High-density epiretinal electrical stimulation	256 electrodes	[20]
Alpha IMS/AMS	Retina Implant AG (Germany)	Subretinal	Subretinal micro-photodiode-based stimulation	IMS: ~1,500 photodiode pixels; AMS: 1,600 photodiode pixels, 70 μm pitch	[10, 24, 25]
PRIMA	Pixium Vision (France)	Subretinal	NIR photovoltaic stimulation	378 pixels; 100 μm pixel size	[12, 21]
Gen2 Bionic Vision	Bionic Vision Technologies (Australia)	Suprachoroidal	Electrical stimulation via the suprachoroidal array	44-channel array	[18, 22]

Another crucial consideration is the relationship between array geometry and the extent of electrode retinal coverage. Existing clinical devices differ markedly in stimulation-site counts, electrode pitch, substrate dimensions, and effective visual field. For instance, Argus II utilizes a 60-electrode epiretinal array, whereas subretinal devices, such as Alpha IMS/AMS, employ denser photodiode-based architectures. Similarly, the PRIMA system leverages a compact photovoltaic pixel array specifically designed to restore central vision in patients with macular disease [10–12]. However, increasing the number of stimulation sites does not necessarily improve functional vision. Visual performance also depends on implant coverage, electrode-retina proximity, stimulation mode, and the spatial selectivity of neural activation [13, 17].

Overall, epiretinal, subretinal, and suprachoroidal strategies represent distinct engineering and biological trade-offs rather than a single optimal design. Epiretinal systems offer relatively straightforward surgical access and direct GC stimulation. Conversely, subretinal implants engage residual retinal processing with potentially improved spatial correspondence, and suprachoroidal systems provide a less invasive implantation route at the expense of a wider gap from target neurons [9, 13, 22]. Placement-related differences are crucial to the interpretation of device performance because the implantation approach itself affects stimulation thresholds, the electrical-activation spread, and effective spatial resolution. Accordingly, decoding the interplay among electrode spacing, array geometry, and reported visual acuity remains central to evaluating current retinal prosthetic systems.

Clinical and regulatory status

While the clinical translation of retinal prosthetic systems has demonstrated that artificial retinal stimulation can restore partial visual function in selected patients with severe photoreceptor degeneration, sustainable commercial success remains elusive. Despite substantial technological advances, only a few systems have achieved regulatory approval, and even fewer have progressed to sustained clinical adoption. Clinical studies consistently report improvements in basic visual functions, such as light perception, object localization, and motion discrimination, alongside performance gains in selected orientation or mobility tasks. However, the level of restored vision remains well below the threshold necessary for high-acuity tasks, such as normal reading or detailed object recognition [11, 17, 23]. The reported clinical outcomes highlight the feasibility and current limitations of retinal prosthetic technologies. Table 2 summarizes differences in clinical milestones, regulatory status, and current development stage.

Table 2.

Clinical and regulatory status of representative retinal prosthetic platforms

Device	Key milestone (year)	Regulatory status	Current development status	References
Argus II	FDA HDE approval (2013); Five-year clinical report (2016)	FDA HDE approval; CE marked	Discontinued commercial production	[11, 19, 26]
IMIE 256	First-in-human report (2021)	Early clinical trial	Long-term adoption and regulatory translation have not been established	[20]
Alpha IMS	Clinical interim report (2015)	CE marked	No widespread global adoption	[10, 17]
Alpha AMS	Multicenter clinical report (2017)	CE marked	No widespread global adoption	[17, 24]
PRIMA	First clinical report (2020); Multicenter clinical update (2026)	Ongoing clinical trial	Active clinical development	[12, 21]
Gen2 Bionic Vision	First-in-human report (2014); Second-generation feasibility report (2025)	Early-phase clinical trial	Under continued development	[18, 22]

Among evaluated epiretinal devices, the Argus II retinal prosthesis system has been the most widely recognized in clinical settings. The system received Conformité Européenne (CE) mark approval in Europe and the Food and Drug Administration (FDA) Humanitarian Device Exemption (HDE) approval in the United States for treating patients with advanced RP [26]. Clinical trials demonstrated that Argus II improved object localization, motion discrimination, and basic orientation performance in profoundly blind individuals, thereby establishing a proof of concept for epiretinal prosthetic vision [11, 19]. Nevertheless, the level of restored vision remained limited, resulting in the subsequent commercial discontinuation of the system, illustrating the gap between technical feasibility and sustained clinical deployment [17].

More recently, newer epiretinal platforms, such as the IMIE 256, were evaluated clinically as next-generation high-density systems. While an initial human trial demonstrated encouraging safety and efficacy over a semichronic implantation period in patients with end-stage RP [20], IMIE 256 remains an early-stage clinical platform because long-term adoption, broader multicenter validation, and definitive regulatory translation have not been established.

For subretinal prostheses, the Alpha IMS/AMS systems represented major milestones in Europe. Notably, Alpha IMS demonstrated that subretinal stimulation restores form vision, supports letter recognition in several patients, and achieves measurable visual acuity in selected cases [10]. Alpha AMS, developed as a successor to Alpha IMS, subsequently received CE mark approval, displaying improved durability relative to its predecessor [24]. Despite these notable advances, neither system progressed to widespread global adoption, indicating that promising clinical findings alone are insufficient to ensure long-term commercial and regulatory success [17].

Regarding AMD, the PRIMA photovoltaic retinal prosthesis has emerged as a next-generation subretinal platform for patients with geographic atrophy secondary to atrophic AMD. Earlier clinical trials demonstrated restored central visual percepts and measurable prosthetic acuity [12]. More recent multicenter clinical data further confirmed that this strategy improved visual acuity over 12 months for patients with AMD-triggered geographic atrophy, supporting the feasibility of PRIMA for the partial restoration of central vision [21]. The PRIMA system remains the most appropriately characterized, still undergoing clinical trials rather than being cleared for routine clinical use.

Similarly, suprachoroidal retinal prostheses have consistently advanced through early-phase clinical trials, although they have not achieved established regulatory approval or routine clinical adoption. First-in-human and second-generation investigations demonstrated acceptable safety profiles along with improved functional vision, basic daily activities, and observer-rated quality of life [17, 22]. These findings indicate that the suprachoroidal approach remains clinically relevant as an alternative retinal prosthetic platform, although it is still best characterized as an early-stage clinical development phase [17, 22].

Collectively, the current clinical and regulatory landscape validates proof of concepts for retinal prosthetic vision but highlights that routine, high-resolution visual restoration remains elusive. As summarized in Table 2, representative retinal prosthetic systems have followed markedly different translational paths, ranging from regulatory approval and active clinical trial status to commercial discontinuation despite foundational scientific success [17, 23]. This mixed outcome indicates that regulatory approval alone does not guarantee real-world clinical success. The real-world adoption of retinal prostheses also depends on functional benefit, surgical feasibility, long-term safety, device durability, rehabilitation burden, and economic sustainability.

Electrode spacing, visual acuity, and the visual field

In retinal prosthesis literature, visual performance is quantified via distinct yet complementary metrics. While visual acuity refers to the capability to resolve fine spatial detail, the visual field defines the extent of the visual area subtended by the implant. Furthermore, established visual acuity lacks a standardized assessment protocol. Some studies reported grating acuity, which reflects the ability to discriminate alternating stripe patterns, whereas others employed Landolt C recognition, which is based on identifying the spatial orientation of a gap within a ring-shaped optotype. Some studies also assessed acuity via the Tumbling E test, a psychophysical task in which subjects identify the orientation of the capital letter E, or related optotype-based tasks that rely on calibrated visual symbols to quantify visual acuity [20]. Additionally, visual acuity may be expressed in Snellen, decimal, or logMAR notation. For example, 20/20 in Snellen notation represents standard visual acuity, whereas 20/400 indicates markedly reduced acuity, meaning that a patient must be stationed 20 feet away to read a target that a person with standard vision could read at 400 feet. Because these disparate testing paradigms and notation systems are not directly equivalent, numerical cross-device comparisons require cautious interpretation. Other reported outcomes, such as light perception, object localization, motion discrimination, and orientation or mobility tasks, reflect broader aspects of functional vision rather than visual acuity itself [17, 19, 20].

Theoretical spatial resolution and retinal coverage

Theoretical spatial resolution in retinal prostheses is typically estimated from electrode or pixel pitch because the spacing between adjacent stimulation sites constrains the capacity of the implant to represent high-resolution spatial detail. Approximately 288 μm on the retina corresponds to 1° of visual angle [19]. In simplified terms, reduced electrode or pixel spacing is expected to support finer spatial resolution, whereas the overall array dimension influences the breadth of the visual field covered by the implant [13, 17, 19].

In addition to electrode spacing, the overall array dimensions are crucial because of their very close relationship with the potential visual field provided by the implant. While electrode or pixel pitch constrains how finely visual information can be sampled, the total retinal area covered by the array determines the overall visual scene accessible to the patient [13, 17]. The distinction between spatial resolution and visual field is crucial, as both factors represent distinct yet complementary aspects of prosthetic vision. Notably, a device may enable relatively fine spatial sampling through a smaller pitch while still providing only a limited field of view if the stimulated retinal area is narrow. Differences between array geometry and retinal coverage are illustrated in Fig. 2, which also compares representative prosthetic systems across stimulation-site configuration and approximate retinal footprint.

Fig. 2.

Comparison of representative retinal prostheses across array configuration and approximate retinal footprint.a Schematic comparison of the nominal retinal footprints of representative prosthetic arrays relative to the fundus. Device outlines are shown with reference to the macula, foveal center, and optic disc to illustrate differences in array size and approximate retinal coverage. The footprints correspond to Argus II (3 × 5 mm; 60 electrodes), IMIE 256 (4.75 × 6.5 mm; 256 electrodes), Alpha implantable micro-photodiode stimulator (Alpha IMS; 3 × 3 mm; ~1,500 pixels), Alpha advanced micro-photodiode stimulator (Alpha AMS; 2.8 × 2.8 mm; 1,600 pixels), and PRIMA (2 × 2 mm; 378 pixels). Scale bar = 5 mm. b–e Representative examples of clinically developed retinal prostheses demonstrating differences in device architecture and physical footprint: b Argus II, c IMIE 256, d Alpha IMS (top) and AMS (bottom), and e PRIMA. Panel-specific source credits: Panel b was adapted from [27], distributed under the terms of the Creative Commons Attribution License, and from [28], licensed under CC BY 3.0; Panel c was adapted from [20], licensed under CC BY 4.0; Panel d was adapted from [29], with permission from John Wiley and Sons, and from [17], with permission from Elsevier; Panel e was adapted from [30], with permission from Elsevier

The relationship between physical pitch and theoretical resolution is well illustrated using representative retinal prosthetic systems. For instance, the Argus II epiretinal prosthesis features an electrode array with an approximately 575 μm center-to-center spacing, which translates to roughly 2° of visual angle per electrode interval. Because grating acuity requires the discrimination of alternating stripe patterns, the theoretical grating limit of Argus II is established at approximately 4° [11, 19]. Subretinal systems, such as Alpha IMS/AMS, employ a much denser stimulation architecture (pixel pitch: 70 μm), which is estimated to correspond to a theoretical visual acuity of approximately 20/250 to 20/280 [16, 25].

However, these theoretical values must only be interpreted as nominal estimates rather than direct predictors of patient vision. Particularly, the Argus II value is expressed as a grating-based visual-angle limit, whereas the Alpha IMS/AMS values are typically expressed as Snellen-equivalent acuity estimates. Although both are derived from device geometry, they are not directly comparable and must not be considered interchangeable. More crucially, prosthetic acuity depends on electrode spacing, as well as electrode size, current spread, charge delivery, and the degree of effective implant-target retinal neuron coupling [8, 13, 17]. Consequently, although smaller spacing may improve theoretical resolution, it does not necessarily translate to proportional improvements in functional visual performance.

The limited visual field covered by current retinal prostheses must be considered when interpreting functional performance, as it remains constrained compared with the expansiveness of natural vision [19, 31]. For example, the visual field of the Argus II spans approximately 11° × 19°, while the Alpha AMS offers a maximum field of approximately 13° and the PRIMA covers roughly 7°. Even higher-density configurations such as the IMIE 256 provide a modest footprint of 20.1° × 14.7° [12, 19, 20, 24]. These examples indicate that implant geometry constrains the fineness of spatial sampling and the extent of the visual scene accessible through prosthetic stimulation.

Reported clinical visual acuity

Reported clinical visual acuity in retinal prostheses remains limited, varying with device architecture and deployed testing paradigm. In the Argus II system, the best reported grating acuity has been summarized as approximately 20/1260, although only a subset of implanted patients reliably completed spatial discrimination tasks [16, 17]. Grating acuity is typically measured by identifying the finest stripe pattern a patient can distinguish, expressed as spatial frequency in cycles per degree (cpd), where one cycle corresponds to one bright and one dark stripe pair within one degree of visual angle. This spatial-frequency value can be converted into an approximate Snellen-equivalent acuity. For example, a Snellen-equivalent acuity of 20/1260 corresponds to a spatial resolution of approximately 0.48 cpd. In practical terms, this means that the patient must be 20 feet from a target that a person with normal vision could identify from 1,260 feet. Although this level of vision is profoundly impaired, compared with normal visual acuity, it represents measurable spatial discrimination beyond simple light perception. Subretinal systems have generally demonstrated slightly better acuity outcomes. In Alpha IMS, measurable acuity, including Landolt C up to 20/546, has been achieved in selected patients [10, 16]. In Alpha AMS, visual acuity measured using Landolt C was reported in only two patients, with values of 20/546 and 20/1111 [24, 25].

More recent retinal prosthetic platforms have yielded measurable acuity outcomes, although these were obtained across different testing paradigms. In the PRIMA photovoltaic system, prosthetic acuity in patients with central implant placement was in the 20/460−20/550 range, whereas one patient with off-center implant placement achieved 20/800 [12]. In the IMIE 256, visual acuity values of 20/800 and 20/1200 were achieved in grating-acuity and Tumbling E tasks, respectively [20]. Compared with normal visual acuity (⁓20/20), which corresponds to a minimum resolution angle of 1 arcmin and an approximate grating resolution of 30 cpd, all reported prosthetic acuity values remain markedly limited. In Snellen-equivalent notation, a smaller denominator indicates improved spatial discrimination. Accordingly, values such as 20/460 − 20/550 indicate better spatial discrimination than 20/1260, although even 20/460 corresponds to a minimum resolvable angle more than 20-fold that of normal vision. Theoretical estimates derived from electrode or pixel spacing assume uniform sampling and independent stimulation channels, whereas clinical acuity is constrained by the current spread, electrode-retina separation, axonal activation, phosphene elongation, and residual-retinal degeneration, alongside subject-specific fitting conditions. Consequently, apparent cross-device differences must be interpreted cautiously, as reported acuity depends on psychophysical testing methodologies, patient-selection criteria, residual-retinal condition, visual-field location, device fitting, and whether the task measures grating detection or standard optotype recognition. Overall, higher-density architectures considerably improve acuity outcomes, although the improvement degree remains below that expected from implant geometry alone.

Because these values are derived from different test paradigms, they must not be interpreted as fully interchangeable even when expressed in similar notation. As grating acuity, Landolt C, and Tumbling E assess different aspects of visual discrimination, direct cross-device numerical comparisons must proceed cautiously [17, 32].

Additionally, established acuity values must be interpreted with consideration for device-related differences in the visual field. A system may demonstrate measurable improvement in acuity while still providing only a restricted field of view, thereby potentially limiting its functional benefit for orientation and mobility [19, 31].

Resolution-gap quantification

The persistent difference between theoretical spatial resolution and actual patient performance may be described as a resolution gap. In retinal prostheses, this metric quantifies the difference between the nominal resolving capability predicted from device geometry and the effective perceptual resolution achieved clinically. As summarized in Table 3, this gap is evident across representative retinal prosthetic systems when comparing nominal geometric parameters with reported clinical visual acuity. In the Argus II system, the 575 μm center-to-center electrode spacing corresponds to approximately 2° of visual angle on the retina. Because a minimum of two adjacent stimulation sites are required to represent one bright-dark grating cycle, this spacing indicates a theoretical limit of approximately 4° per cycle [11, 19]. Although this geometry-based value can be approximately expressed as 0.25 cpd, corresponding to a Snellen-equivalent acuity of 20/2400, such conversion should be interpreted cautiously. Although the best reported grating acuity for Argus II was 20/1260, a substantial proportion of patients did not attain measurable grating visual acuity [17, 19]. This device-specific example illustrates that the nominal sampling limit derived from electrode pitch does not directly predict the functional acuity measured in patients. In Alpha IMS/AMS, the much smaller pixel pitch corresponds to finer theoretical resolution, although the measured clinical acuity remains markedly below that expectation [10, 16, 24, 25]. Similar discrepancies are also apparent in more recent photovoltaic systems, where measurable visual improvements were achieved even though they still fall short of the device-geometry-predicted nominal resolution [12, 21].

Table 3.

Reported geometric parameters, visual field, and visual acuity of representative retinal prosthetic platforms

Device	Array dimensions	Electrode/pixel specification	Reported visual field	Reported visual acuity	Test method	References
Argus II	3 mm × 5 mm	60 electrodes; 200 μm diameter; 575 μm spacing	11° × 19° (≈ 20° diagonal)	~ 20/1260	Grating acuity	[11, 17, 19]
IMIE 256	4.75 mm × 6.50 mm; 70 μm thickness	248 large electrodes (210 μm diameter, 350 μm pitch); 8 small electrodes (160 μm diameter, 300 μm pitch)	20.1° × 14.7°	20/800; 20/1200	Grating acuity; Tumbling E	[20]
Alpha IMS	3 mm × 3 mm	~ 1,500 photodiode pixels; 70 μm pitch	Up to 15° diagonal	20/546	Landolt C	[10, 16]
Alpha AMS	2.8 mm × 2.8 mm	1600 pixels (40 × 40); 70 μm pitch	maximum ~ 13°	20/546; 20/1111	Landolt C	[24, 25]
PRIMA	2 mm × 2 mm; 30 μm thickness	378 hexagonal pixels; 100 μm pixel size/pitch	~ 7° central field	20/460−20/550	Landolt C	[12, 21]

Crucially, the resolution gap does not simply reflect a numerical mismatch between theoretical estimates and experimental outcomes. Rather, it reflects a fundamental limitation of current retinal prosthetic systems. Previous studies demonstrated that the effective spatial resolution of prosthetic vision is influenced by array pitch, as well as the current spread within retinal tissue, electrode-retina distance, unintended activation of neighboring or passing neural elements, charge-injection limits, and degeneration-associated inner-retina remodeling [7, 13–15, 17].

In Table 3, electrode or pixel pitch denotes the center-to-center distance between adjacent stimulation sites, with smaller pitch values indicating a denser nominal stimulation architecture informed by the close packing of electrodes or pixels. Therefore, electrode spacing must be treated as an important yet imperfect predictor of prosthetic visual acuity. As presented in Table 3, devices with smaller pitch or denser stimulation architectures do not necessarily achieve proportionally better reported clinical acuity. Device geometry helps define the nominal resolution of an implant, whereas the actual visual detail perceived by the patient depends on additional biological and biophysical factors, transcending array pitch. The gap between theoretical and clinical resolutions reflects the insufficiency of increasing implant density alone. A more complete description requires considering the biological and biophysical constraints that limit the effective spatial resolution in retinal prosthetic systems.

Factors limiting spatial resolution

To elucidate the resolution gap, it is necessary to examine how specific biophysical constraints degrade the effective independence of stimulation sites. The discrepancy between device-geometry-predicted theoretical resolution and clinically achieved visual acuity suggests that array pitch alone does not sufficiently determine prosthetic performance. Even when stimulation sites are densified, the effective spatial resolution of a retinal prosthesis still depends on how selectively and efficiently electrical stimulation can be delivered to the target retinal neurons. In practice, stimulation selectivity is constrained by multiple interacting factors, including electrode size, current spread within the retinal tissue, and the implant-neural target distance [8, 13, 17]. Rather than influencing spatial resolution independently, these parameters synergistically dictate the stimulation threshold, spatial extent of neural activation, and the reproducibility of the evoked retinal response. For example, increased electrode-retina distance requires higher stimulation amplitude, which broadens current spread and reduces the spatial independence of neighboring stimulation sites.

Such interactions determine whether neighboring electrodes function as independent visual-information channels. Figure 3 provides a conceptual summary of the major factors accounting for the gap between nominal device resolution and effective clinical acuity. The figure represents an overview of contributing limitations rather than a mechanistic model describing all interactions among electrode geometry, current spread, electrode-retina coupling, and retinal degeneration.

Fig. 3.

Conceptual overview of factors limiting effective spatial resolution in retinal prostheses. Electrode size, current spread, electrode distance from the retina, retinal remodeling, and unintended axonal stimulation may reduce spatial selectivity and contribute to the gap between theoretical resolution and clinical visual acuity

Electrode size and electrical efficiency

Electrode size is closely related to spatial selectivity and electrical efficiency in retinal stimulation. Generally, smaller electrodes are attractive for their ability to confine stimulation to a more localized retinal region, potentially improving spatial resolution. The concept of local confinement has driven the development of higher-density arrays across epiretinal and subretinal prosthetic systems [13, 17].

However, reducing electrode size does not provide an unlimited advantage. Decreasing the electrode area limits the charge density that can be delivered safely through the electrode owing to the reduced available electrochemical surface area. This reduced deliverable charge primarily reflects charge-injection capacity and electrochemical safety at the electrode surface rather than direct cell damage. Concurrently, electrode size influences the geometric confinement of stimulation, as well as the threshold, dynamic range, and target-neuron recruitment [8, 13]. A recent study on high-acuity electrode design illustrated the trade-off between electrode miniaturization and safe charge delivery more directly. Damle et al. compared 10, 20, and 30 μm planar iridium oxide (IrO₂) electrodes at a 50 μm pitch, demonstrating that smaller-diameter electrodes exhibited lower stimulation thresholds, whereas larger-diameter electrodes provided a broader, usable dynamic range within safe charge-injection limits while recruiting more GCs [33]. These findings indicate that a lower threshold alone cannot sufficiently define the optimal electrode size because larger-diameter electrodes may support safer or more reliable charge delivery over a wider operating range. Consequently, there exists a practical lower limit to planar electrode size, below which geometric miniaturization may fail to translate into improved functional stimulation performance.

The implication of the trade-off between miniaturization and safe charge delivery also depends on electrode-material properties. Charge-injection capacity and electrochemical stability differ across typically deployed materials, such as platinum and IrO₂, with material-dependent electrochemical differences becoming increasingly prevalent as electrode size decreases [8, 13]. Electrode materials with higher charge-injection capacity, such as activated IrO₂, better support smaller stimulation sites by enabling sufficient charge delivery within safe electrochemical limits. Conversely, materials with lower charge-injection capacity may restrict electrode miniaturization because the required stimulation charge may approach safety limits with the decrease in the electrode area. Therefore, the practical limit for electrode miniaturization is determined by electrode geometry and the electrochemical properties of the electrode material at the electrode-tissue interface. In this context, advanced material strategies encompass approaches such as the development of high-charge-injection electrode materials, surface modifications that increase the effective surface area, and the integration of conductive, mechanically compliant materials into electrode arrays to improve charge transfer and long-term stability [34].

The trade-off between geometric miniaturization and safe, reliable charge delivery becomes even more relevant in degenerated retina, where stimulation thresholds may already be elevated via retinal remodeling, altered excitability, or imperfect implant-tissue coupling [7, 15, 17]. Under such conditions, very small electrodes may still require very high stimulation currents to reliably activate target neurons. However, increasing stimulation current may also enlarge the neural-activation area, thereby offsetting the expected spatial-selectivity gain. Therefore, simply reducing the electrode size does not guarantee finer functional resolution.

Accordingly, electrode size must be evaluated as a geometric design parameter and as part of a broader balance involving spatial selectivity, stimulation threshold, charge-injection capacity, and long-term interfacial stability. Therefore, the practical bioengineering challenge shifts from simply minimizing electrode dimensions to identifying an optimal range of electrode size and material combination that simultaneously supports localized stimulation and reliable neural activation.

Current spread within the retinal tissue

Current spread within retinal tissue represents a primary factor limiting effective spatial resolution in retinal prostheses. Although an implant array may contain high-density, closely packed stimulation sites, the electric field generated by a single electrode does not remain confined to its geometric boundary. Rather, electrical current spreads through the conductive retinal environment, potentially activating neurons beyond the intended target region [13, 31].

As electrode spacing decreases, the spatial separation of activation produced by neighboring electrodes becomes increasingly important. A smaller center-to-center distance between electrodes increases the nominal density of sampling sites, but improved spatial resolution requires that each electrode activate a retinal region that is sufficiently confined relative to the spacing between adjacent electrodes. The extent of this effective activation field is determined by current spread within the retinal tissue and the stimulation amplitude required to reach threshold. If the activation field of one electrode overlaps with that of neighboring electrodes, adjacent stimulation sites cannot function as independent visual channels. Consequently, reducing electrode pitch or increasing electrode number may increase the geometric sampling density of the array without increasing the number of distinct perceptual units available to the patient. The functional benefit of a denser array therefore depends on the ability of adjacent electrodes to produce spatially separable patterns of retinal activation. Experimental studies have shown that the spatial extent of activation is influenced by stimulation depth, electrode location relative to retinal layers, stimulus amplitude, and the balance between direct neuronal activation and network-mediated recruitment of residual retinal circuitry [14].

In epiretinal stimulation, current spread is further complicated by the unintended activation of passing GC axons. Rather than producing a focal percept, such activation may generate elongated or distorted phosphenes, thereby weakening the correspondence between electrode position and perceived visual location [13, 17]. A more recent experimental study revealed that the lateral spread of electrical activation remains a major obstacle to achieving focal stimulation, even in higher-density designs [31].

Collectively, these findings indicate that the current spread is not merely a secondary technical issue; rather, it represents a central biophysical constraint of retinal prosthetic systems. It reduces the functional independence of neighboring stimulation sites and therefore contributes directly to the gap between theoretical and clinical resolution.

Electrode-retina distance and interface stability

The distance between the stimulating electrode and target retinal neurons represents another key factor limiting effective spatial resolution in retinal prostheses. Generally, a shorter electrode-retina distance improves electrical coupling, lowers stimulation thresholds, and enables more localized charge delivery to the intended neural population [13, 17]. Conversely, an increase in the electrode-retina distance must be accompanied by a larger stimulation-current amplitude to reach the threshold, consequently broadening the resulting neural-activation area. Thus, neighboring stimulation sites will become less functionally independent, reducing spatial selectivity.

The criticality of the electrode-retina distance has been demonstrated clinically and via computational and modeling studies. In Argus II subjects, perceptual threshold was associated with electrode-retina distance and related interfacial properties, indicating that even small apposition variations can influence stimulation efficiency [35, 36]. More recent patient-specific computational and modeling studies have also demonstrated that the electrode-retina distance and impedance contribute significantly to threshold variability [37, 38]. Collectively, the clinical and modeling studies cited above indicate the dependence of spatial resolution on array geometry and the quality of the electrode-retina interface. Particularly, a shorter, more stable electrode-retina distance lowers stimulation thresholds and confines neural activation more locally, whereas increased or variable separation from the target retinal layer requires larger current amplitudes, broadens the activated retinal region, and reduces spatial selectivity.

Differences in the electrode-retina distance and interfacial stability also explain the divergence of implantation approaches regarding stimulation efficiency and spatial selectivity. While epiretinal devices are positioned close to the GC layer, the actual interface may vary because of retinal curvature, fibrosis, or a chronic mechanical mismatch between the array and retinal surface. Although subretinal devices are typically placed closer to bipolar-cell-level circuitry, their performance still depends on precise placement and preservation of the residual retinal structure. In suprachoroidal systems, a higher separation from the neural retina is generally associated with increased thresholds and lower spatial selectivity, even when surgical accessibility and safety are favorable [17, 22].

Electrode-retina distance is also shaped by the physical properties of the implant substrate. Notably, substrate thickness, flexibility, and elastic conformity dictate how well the array follows the curvature of the retinal surface. Because many prosthetic arrays are fabricated on relatively planar substrates despite the retina being curved, incomplete conformity can leave portions of the array separated from the tissue, thereby increasing the effective electrode-retina distance. Substrate conformity is particularly relevant in chronic implantation, where even small gaps may influence threshold and spatial selectivity [8, 13].

Fixation strategy is also critical for maintaining electrode-retina coupling. While epiretinal devices often require mechanical fixation to sustain contact with the internal retinal surface, postoperative fibrotic tissue reactions or device micromotion may alter the interface over time even when initial apposition is satisfactory [11, 39]. In subretinal systems, placement accuracy and structural stability are similarly crucial, particularly because local displacement may alter the functional relationship between stimulation sites and the preserved retinal circuitry [10]. Thus, the electrode-retina distance must be considered a purely static geometric parameter, as well as the outcome of a dynamic interface shaped by the substrate design, fixation method, tissue reaction, and chronic anatomical alteration.

From a placement-specific perspective, the major implantation approaches are summarized as follows. Epiretinal systems, such as Argus II and IMIE 256, locate the electrode array adjacent to the GC layer, although the electrode-retina interface may be compromised by retinal curvature, fixation instability, fibrosis, and chronic micromotion. Subretinal systems, including Alpha IMS/AMS and PRIMA, interface more directly with bipolar-cell-level circuitry and may better preserve physiological retinal signal processing. However, their performance depends strongly on accurate placement, conformal contact with the retinal surface, and the preservation of the residual retinal structure. Suprachoroidal systems, exemplified by the Gen2 Bionic Vision platform, are implanted farther from the neural retina, potentially offering surgical and safety advantages; however, the larger electrode-retina distance generally requires greater stimulation amplitudes and results in broader neural activation, thereby reducing spatial selectivity. Overall, each implantation approach represents a distinct trade-off among electrode-retina distance, mechanical stability, surgical accessibility, and achievable spatial selectivity.

Stimulation parameters and perceptual outcomes

While electrode geometry and electrode-retina coupling define the structural constraints of retinal prosthetic systems, the percept ultimately experienced by the patient also depends strongly on how electrical stimulation is delivered. Even when the same array is utilized, changes in waveform, pulse duration, amplitude, and stimulation frequency can alter neural recruitment, thereby altering percept quality, stability, and spatial characteristics [13, 14, 17]. Accordingly, stimulation parameters must not be regarded as mere adjustable device settings but as central determinants of functional visual experience.

For example, clinical evaluations of the Argus II demonstrate that modulating pulse rate and phase duration directly alters brightness saturation and perceptual fading kinetics [11, 19]. Conversely, investigations using the IMIE 256 array highlight discrete phosphene generation during threshold testing via cathodic-first biphasic pulses tailored to a 0.45 ms/phase duration [20]. As summarized in Table 4, representative retinal prosthetic systems utilize highly divergent stimulation paradigms, varying widely in waveform morphology, current amplitude or optical irradiance, phase duration, and stimulation frequency [11, 12, 19–21, 24, 25]. These stimulation variables directly dictate critical perceptual features, such as phosphene elicitation, brightness dynamics, temporal persistence, threshold responses, and spatial focality [13, 14, 17]. For example, clinical evaluations of the Argus II demonstrate that pulse rate and phase duration directly affect phosphene brightness dynamics and perceptual fading kinetics [11, 19], whereas IMIE 256 demonstrated distinct phosphene elicitation during threshold testing via cathodic-first biphasic pulses tailored to a 0.45 ms/phase duration [20]. In subretinal systems, Alpha AMS relies on current- and voltage-controlled biphasic stimulation tailored to device-specific phase durations [24, 25]. Conversely, the PRIMA system leverages NIR photovoltaic stimulation to generate localized retinal signals under device-specific irradiance, phase-duration, and frequency conditions [12, 21]. Collectively, the stimulation-related observations summarized in Table 4 indicate that device structure alone perceptual outcomes cannot be interpreted solely from the device structure. Stimulation waveform, amplitude, phase duration, and frequency must be considered together with electrode placement, array geometry, and residual retinal interface when comparing retinal prosthetic platforms.

Table 4.

Representative stimulation parameters reported for selected retinal prosthetic platforms

Device	Pulse type	Stimulation amplitude (current amplitude or optical irradiance)	Phase duration (ms/phase)	Stimulation frequency	Stimulation-related findings	References
Argus II	Cathodic-first biphasic current pulses	4-677 µA	0.46 or 0.97	3–60 Hz	Pulse duration affected phosphene brightness dynamics and fading kinetics.	[11, 19]
IMIE 256	Cathodic-first biphasic current pulses	Threshold current: 236.6 µA (large electrodes; range, 33–400 µA); 241.7 µA (small electrodes; range, 88–400 µA)	0.45	1 Hz threshold testing; 30 Hz functional stimulation	0.45 ms/phase threshold testing elicited distinct phosphenes.	[20]
Alpha AMS	Cathodic-first biphasic, current- and voltage-controlled pulses	Up to 50 µA	0.1–2.0	1–10 Hz	Biphasic stimulation elicited visual percepts and object-related performance.	[24, 25]
PRIMA	Biphasic current waveform generated by NIR photovoltaic stimulation	3.0-3.5 mW/mm2 (irradiance)	0.7–9.8	3–60 Hz	NIR photovoltaic stimulation elicited central prosthetic percepts.	[12, 21]

Pulse waveform

Pulse waveform is a fundamental determinant of charge delivery and neural activation in retinal prosthetic stimulation. Most current retinal prosthetic systems utilize biphasic pulses, typically within a cathodic-first configuration, as this design supports effective neural stimulation while minimizing irreversible electrochemical reactions at the electrode-tissue interface [8, 13]. While conventional electrically driven systems, such as the Argus II, Alpha AMS, and IMIE 256 platforms, rely on charge-balanced biphasic stimulation, the PRIMA system utilizes a wireless photovoltaic strategy wherein patterned NIR light is converted into localized, biphasic stimulation currents at the implant [12, 21]. Thus, although the platforms differ in hardware design, they all rely on temporally structured charge delivery to generate reproducible percepts.

Beyond standard implementations, the waveform design is critical for electrochemical safety and the determination of how stimulation is distributed across retinal circuitry. The timing and polarity of the stimulus pulse determine which retinal elements are preferentially activated and how reliably phosphenes are elicited. Although established clinical platforms predominantly implement charge-balanced biphasic waveforms, experimental retinal stimulation studies have investigated asymmetric biphasic pulses, such as those featuring a short high-amplitude primary phase, followed by a longer low-amplitude charge-balancing phase, as a pathway to modulate activation thresholds and cell-type selectivity while maintaining net charge balance [13, 14].

Additionally, leading-phase polarity, including cathodic-first versus anodic-first stimulation, influences neural recruitment and spatial selectivity by altering the relative contribution of direct RGC activation, axonal activation, and network-mediated responses. Because the unintended activation of passing axons elongates phosphenes and reduces perceptual focality, waveform asymmetry and leading-phase polarity represent crucial parameters for optimizing spatially selective retinal stimulation.

Pulse duration and threshold

Pulse duration is closely related to the stimulation threshold and patterns of neural recruitment. Generally, longer pulse durations reduce the current amplitude necessary to elicit a percept, whereas shorter pulses typically require higher amplitudes to reach the threshold [13, 14]. However, the effect of pulse duration is not exclusively limited to the stimulation threshold. Changes in the pulse duration may shift the balance between direct GC activation and indirect residual retinal network recruitment, and this difference can influence perceptual quality and spatial selectivity [14]. The distinction between direct and network-mediated activation is critical because neither activation mode is uniformly advantageous under all conditions. More direct GC activation may enable faster and more immediate responses, although it may also increase the likelihood of unintended axonal activation in epiretinal stimulation. Conversely, a higher recruitment of the residual retinal network may engage inner retinal processing more effectively, although it may also broaden the spatial or temporal response spread. Therefore, pulse duration must be treated as a timing parameter, as well as a factor influencing the mode of neural activation and the resulting perceptual characteristics.

For biphasic electrical stimulation, phase duration refers to the duration of each individual cathodic or anodic phase. Reported phase-duration ranges differ according to array placement and stimulation mechanism. For example, Argus II relies on balanced cathodic-first biphasic pulses with phase duration of 0.46 or 0.97 ms/phase, whereas IMIE 256 utilizes cathodic-first biphasic pulses with a 0.45 ms/phase duration during threshold testing [11, 19, 20]. Alpha AMS utilizes current- and voltage-controlled biphasic stimulation with phase durations ranging from 0.1 to 2.0 ms/phase [24, 25]. PRIMA utilizes a wider temporal range (0.7–9.8 ms/phase) under its photovoltaic stimulation protocol, with phase determined by the duration of the NIR light pulse [12, 21]. Thus, rather than representing a single, universally optimal pulse duration, the reported conditions indicate that the phase duration is selected in a device-specific manner according to the electrode placement, target retinal circuitry, and the underlying charge-delivery mechanism.

Stimulation frequency

Stimulation frequency primarily influences the temporal characteristics of prosthetic perception rather than spatial resolution itself. In retinal prostheses, frequency determines how often stimulation pulses are repeated over time; thus, it dictates whether a percept appears intermittent, sustained, or temporally fused. Changes in stimulation frequency can alter percept duration and temporal discriminability, that is, how long an electrically evoked percept remains visible and how clearly repeated stimulation events can be distinguished [13, 17, 19, 34].

Temporal continuity is crucial because a percept must typically persist long enough to be interpreted by the patient. If stimulation frequency is too low, perceptual continuity may become insufficient, causing the visual pattern to appear unstable or fragmented. Conversely, if the stimulation frequency is too high, repeated stimulation events may become less distinguishable or contribute to fading. Accordingly, stimulation frequency is closely related to the temporal persistence and continuity of prosthetic vision, although these factors may also exert a secondary influence on pattern recognition [13, 17].

The frequency ranges utilized across representative retinal prosthetic systems vary substantially. For instance, Argus II and PRIMA generally operate within the 3–60 Hz range, whereas Alpha AMS has been reported to use a broader range (1–10 Hz). Additionally, IMIE 256 permits low-frequency stimulation, beginning at 1 Hz. The device-specific frequency ranges indicate that there is no single frequency range that can be generalized across all retinal prosthetic platforms. Rather, the effective range depends on the specific device architecture, retinal target, and intended perceptual outcome [12, 20, 21, 24].

Accordingly, stimulation frequency must be considered a parameter that shapes the temporal stability, persistence, and continuity of prosthetic perception rather than as a simple timing setting. Its implication lies in balancing percept persistence and temporal distinctness under different stimulation conditions.

Discussion

The central limitation of contemporary retinal prostheses is not a deficit in electrode density; rather, it is the insufficient functional independence of stimulation sites. Current evidence indicates that the performance of retinal prosthetic systems is determined by multiple interacting design and biological factors, rather than any single device parameter. Although electrode spacing is typically employed as a convenient indicator of theoretical resolution, clinical outcome depends on a broader combination of factors, including electrode size, array pitch, retinal coverage, current spread, electrode-retina distance, and stimulation strategy [13, 14, 17, 34].

Consequently, improving prosthetic vision must not be approached as a simple issue of increasing array density. A smaller pitch may increase the nominal number of stimulation sites, whereas the functional value of that increase depends on whether adjacent electrodes activate sufficiently distinct retinal populations. Electrode size is also crucial because it affects geometric selectivity, as well as stimulation threshold, charge-injection limits, and the usable dynamic range [8, 33]. Concurrently, the total retinal coverage of the array remains clinically relevant because prosthetic performance depends on spatial resolution and the extent of visual field available to the patient. Therefore, future optimization must consider electrode size, spacing, and retinal coverage together rather than treating them as independent, isolated design variables.

The electrode-retina interface represents another major determinant of clinical performance. Even with a favorable array geometry, functional resolution may remain limited if the implant does not maintain stable, close coupling to the target retinal tissue. Variations in apposition, fibrosis, chronic micromotion, and geometric mismatch between a relatively planar array and the curved retinal surface may influence stimulation threshold, current spread, and perceptual stability [26, 35, 37, 38]. For chronic retinal implants, substrate flexibility, thickness, and the fixation strategy do not represent secondary engineering considerations; rather, they are critical determinants of long-term stimulation performance. Interfacial stability is particularly crucial for chronic implantation, where an initially favorable interface may evolve.

The stimulation strategy represents another major determinant of prosthetic performance. As established, pulse duration, waveform, and frequency influence not only threshold, but also the mode of neural recruitment and the temporal stability of the percept [13, 14, 34]. Consequently, even when utilizing the same array, distinct perceptual outcomes may emerge depending on how charge is delivered and which retinal elements are recruited. Thus, future improvement in prosthetic vision will require the co-optimization of device structure and stimulation parameters rather than treating hardware and stimulation as independent design domains.

The preceding design constraints also highlight the importance of electrode and substrate material selections. Materials exhibiting higher charge-injection capacity and favorable electrochemical stability may better support miniaturized electrodes, whereas substrate materials with appropriate mechanical flexibility and durability may enhance anatomical conformity to the curved retinal surface and long-term interface stability [8, 34]. Therefore, material selection is not merely a variable of fabrication convenience; it directly dictates whether theoretically favorable array designs can be translated into safe, stable, and functionally effective chronic implants.

Another hurdle in the field is that clinical interpretation remains limited by heterogeneous outcome measures. Reported visual acuity values are derived from highly divergent testing paradigms, ranging from grating acuity and Landolt C to Tumbling E and various task-based assessments. These measures are not fully interchangeable, even when normalized in similar notation systems [32, 40, 41]. Additionally, numerous studies emphasize broader functional outcomes, such as object localization, motion discrimination, mobility performance, or daily basic activities. While these measures are clinically meaningful, they complicate direct comparison across devices. Thus, developing more standardized, well-defined outcome measures would improve the comparability of clinical studies and provide more valuable feedback for device design and optimization.

Several implications can be drawn from these considerations. First, next-generation devices must be developed by more closely integrating array geometry, interface engineering, and stimulation strategy. Second, resolution enhancements must be evaluated not merely by nominal design parameters, but by the long-term stability and reproducibility of percept formation. Third, device performance must be validated using outcome measures that distinguish spatial resolution, temporal characteristics, and functional vision. Implementing these assessment steps would ease the determination of which design modifications yield meaningful clinical benefit.

In summary, advancing the clinical efficacy of retinal prosthetic systems requires coordinated innovations across multiple interdependent domains, including electrode size, spacing, and retinal coverage; control of the current spread; stable electrode-retina coupling; the stimulation strategy; and the material and mechanical properties selected for safe chronic implantation. Next-generation devices must simultaneously enhance the fineness of spatial sampling while expanding the visual field, maximizing the selectivity of neural activation, and preserving the long-term stability of the implant-tissue interface. Therefore, the central challenge is in translating multiparametric breakthroughs in array design, interface engineering, and stimulation control into spatially independent, information-rich percepts that meaningfully extend visual function.

Author contributions

Conceptualization: Seoyoung Hwang, Hee Soo Jeong, Sang Beom Jun. Literature review and data curation: Seoyoung Hwang, Hee Soo Jeong. Writing-original draft preparation: Seoyoung Hwang. Writing-review and editing: Hee Soo Jeong and Sang Beom Jun. Supervision: Sang Beom Jun. All authors read and approved the final manuscript.

Funding

Open Access funding enabled and organized by Ewha Womans University. The authors received no specific funding for this work.

Declarations

Conflict of interest

The authors declare no conflict of interest.

Ethical approval

This article does not contain any studies with human participants or animals performed by any of the authors.

Consent to participate

Not applicable.

Consent to publish

Not applicable.